Glitches are charged particle hits. In the ``best'' cases a single pixel in a single frame shows the effect of the hit. In the worst cases up to 120 pixels are affected in a single frame or the glitch introduces a transient which lasts for several readout and/or changes in the pixel sensitivity. The mean glitch frequency is ~1 glitch per second and on average 8 pixels are affected by a single hit.
The 'mm' option in the CIA procedure deglitch is currently the best choice for automated deglitching. This algorithm employs median filtering (with varying filter sizes) to identify and mask glitches. The method has been known to fail for glitches which last close to or longer than the duration of the observing mode. Additionally, the deglitcher does not mask or correct glitch induced transients (next section). The CVF data cubes or rasters with very strong point sources are generally problematic for automated routines. One last warning about automated deglitchers: they can sometimes confuse sources with glitches (especially strong spectral features in the CVF data cubes).
If all auto-mated options fail. The CIA contributed tool deglitch_view can be used to identify and mask glitches by eye. This is a slow, tiresome process but it is recommended that you should verify the performance of any auto-mated deglitcher manually (in particular for the CVF data cubes).
A word of caution about manual deglitching: If manual deglitching is applied to every raster position image of a raster map, one soon finds that the "depth" of the manual deglitching is hard to gauge -- it varies throughout the cube -- and one never knows when to stop. Eventually, even the fain but real sources start to be cut out. To quote one anonymous CAM expert:
"Be Carefull with manual deglitching - it leads to self-absorbtion, and eventually you go blind!"
Recommended reading:
Expected time commitment: